Study On Vascular Segmentation And Cardiovascular Risk Factor Prediction Algorithm Of Retinal Image Based On Deep Learning

The retinal blood vessel,as the only internal vascular system that can be observed in the human body under non-invasive conditions,has been widely concerned by researchers.Retinal images contain rich information related to diseases.Many important systemic diseases of the human body will cause specific reactions on the retinal image.Therefore,retinal images are widely used in the field of medical aided diagnosis.A large number of clinical experiments have shown that diabetic retinopathy is related to abnormalities of retinal arteries and veins.In addition,hypertension and some pancreatic diseases can cause abnormalities in retinal blood vessels.Quantitative and qualitative analysis of retinal blood vessels plays a very important role in the prevention,diagnosis,monitoring and management of many diseases,especially chronic diseases.Therefore,automatic analysis of retinal images and accurate segmentation of retinal blood vessels and arteriovenous veins are of great significance for disease analysis and diagnosis.Due to the complex structure of retinal blood vessel images,manual analysis by medical personnel such as ophthalmologists is a time-consuming and expensive task.Therefore,many retinal blood vessel and arteriovenous segmentation methods based on image processing and deep learning have been proposed,including fully automatic and semi-automatic methods.However,most algorithms have the problem of retinal arteriovenous segmentation discontinuity.That is,the blood vessel segments are not connected,or the same type of blood vessels contains different types of blood vessels,which makes them difficult to apply directly to clinical data.Based on this,this paper proposes that the loss of topological structure restricts the continuity of blood vessels.The main contents of this article are as follows:First,this article briefly describes the research background,current status and significance of retinal blood vessels.Secondly,this article briefly introduces the retinal blood vessel and arteriovenous segmentation process,the related technology of image processing and the application of convolutional neural network on retinal images.Further,for the difficulty of segmentation of small blood vessels,we proposed a retinal blood vessel segmentation model based on attention mechanism.On this basis,for the problem of discontinuous segmentation of retinal arteriovenous,the topological structure is introduced toconstrain it.At the same time,combined with the generative adversarial networks and topological structure constraints,to achieve automatic segmentation of retinal arteriovenous.Finally,we apply the model to the realistic scene of cardiovascular risk factor prediction based on retinal images,with a view to discovering clinical markers that can be used to prevent and diagnose diseases.This paper uses the data collected by the Guangdong Eye Institute,Department of Ophthalmology,Guangdong Provincial People’s Hospital Ophthalmology Center of the First People’s Hospital of Guangdong Province,and the two public databases DRIVE and STARE.The analysis and comparison with the existing methods prove that our proposed method is robust,effective and accurate.